CONSUMER AND MARKET DEMAND AGRICULTURAL POLICY RESEARCH NETWORK Modelling Functional Food Choice and Health Care Impacts: A Literature Review PROJECT REPORT Modelling Functional Food Choice and Health Care Impacts: A Literature Review

The global market for functional foods is estimated to be worth about US$33 billion For individual consumer choice, the classical consumer theory of utility maximizing might be the most useful framework; we also need to consider the effects of information asymmetry and uncertainty, labelling information, and the effects of consumption decisions of perceived health outcomes. What causes consumers to under-consume healthy food or over-consume unhealthy food? Consumers face two types of uncertainty: uncertainty about the health 1 attributes of a specific food (does this food contain elevated levels of conjugated linoleic acid (CLA)?), and uncertainty over future health outcomes (will consumption of foods high in CLA improve my long-term health, and to what extent?). Information asymmetry therefore underlies consumers' decision regarding the consumption of functional foods. Labelling can play an important role in informing consumers: if the label is credible, and if consumers know how to use the information on the label. Given the importance of labelling and information asymmetry to understanding the consumer decision-making process for functional foods, this paper discuss examples of how these issues have been explored in the literature. The paper focuses on three modified choice methods that represent very different ways of modelling consumer preferences for functional food: the Stated Preference Choice Model with Discrete Choice Analysis, Dependent Preference Model, and modified Protection Motivation Theory. For each method, the paper provides the underlying reasons why this method might be useful to model functional food, presents the theoretical framework underlying each approach, and discusses the advantages and limitations of the method. The paper is organised as follows: section 2 discusses some key characteristics of the emerging functional food sector and discusses the information asymmetry problem facing consumers. Section 3 provides a detailed discussion of three alternative consumer choice models. The final section presents conclusions and suggestions for future research directions. Functional Food and Information Asymmetry What are functional foods? The central idea is that they are foods with health benefits. There are a number of definitions; for example, Poulsen (1999) defines functional food as: "A product which has been modified or enriched with naturally occurring substances with a specific physiological preventative and/or health-boosting effect. The product must also be part of the normal daily consumption of food/fluids." (p.1). Poulsen (1999) also categorizes four 'enrichment' methods by which functional foods can be produced: upgrading, by adding more of a substance that the food already 2 contains; substitution, by replacing a substance with a similar but healthier component; enrichment, by adding an ingredient not normally found in the food product; and elimination, by removal of an unhealthy ingredient from a food. Health 1 In 2003, new regulations were introduced in Canada allowing health claims for Natural Health Products (herbal remedies, homeopathic medicines, vitamins, minerals, traditional medicines, probiotics, amino acids and essential fatty acids), but not for functional foods. 4 The market for functional foods is characterized by information asymmetry for consumers with respect to the presence, level and efficacy of the functional attributes. This point has been recognised in the literature: Menrad et al In many cases, information about food quality can be considered non-rival and nonexcludable; substitutes (for the information) do not exist and if made available to one consumer, the information is usually available to all others. In this respect, quality information could be viewed as a public good. Akerlof's (1970) well-known paper on the market for lemons underpins most subsequent work on information asymmetry as a cause of market failure. Briefly, he argued that good quality products could be eventually driven out of the market by poor 5 quality products because of information asymmetry between buyers and sellers. Sellers have more knowledge about product quality; if the quality cannot be credibly signalled, goods of both qualities will sell at the same price. The absence of a price premium for higher quality results in only low quality products being offered for sale. Two information problems are apparent in the functional foods market. One is whether the functional component is present, such as the presence of elevated levels of omega-3 fatty acids. Of course, this is related to different quality levels of functional foods: higher quality (functional foods) and lower quality (no specific health benefit: 'normal' foods). The other information problem is whether the functional health claim is credible, which depends on whether consumers believe that consumption of functional foods will lead to specific health outcomes. If consumers cannot identify functional foods, they will not be willing to pay a higher price for foods with those qualities, and the economic incentive to produce functional food is not present. Eventually there would be no functional food in the market. Clearly, producers of functional food have a strong incentive to label their products. Labelling can be viewed as a means of providing information for consumers to make choices, and there is a lot debate about the merits of different labelling policies. Compared to the labelling of functional food, the labelling of GM foods has been studied much more widely. The GM food labelling literature provides some useful insights for potential research directions with respect to functional foods. Hu et al Modelling Consumer Choice of Functional Food The last section explained how information plays a key role when consumers make food choices. In this section we take a closer look at the consumer choice process and how consumer decisions can be modelled. The basic economic framework of individual preferences is the standard microeconomic consumer theory of maximizing utility. An individual consumer chooses a consumption bundle faced with his budget restriction. It is assumed that the consumer will exhibit rational behaviour: choosing the bundle which is at least as good as any other among all the bundles. Manski's (1977) random utility approach is more consistent with consumer theory. The individual consumer is always assumed to select the alternative with the maximum utility. However, the analyst does not know the utilities with certainty, and must take the utilities as the random variables, known as random utility. Usually the random utility can be separated into deterministic (observed) and random (unobserved) components for the utility function. Three different consumer choice models are examined in this section: the Stated Preference Model with Discrete Choice Analysis, Dependent Preference Model, and modified Protection Motivation Theory. Stated Preference Model with Discrete Choice Analysis The Stated Preference Method is a useful empirical research technique to understand and predict the decision makers' choice behavior among discrete products. This method There are two kinds of choice data: Stated Preference (SP) and Revealed Preference (RP). The SP data are generated from the decision experiment (survey), while the RP data are from consumers' actual observed choice decisions. There are some major advantages of a stated preference method compared with revealed preference studies: for example, the SP method allows researchers to estimate and predict the demand for new products with new attributes; in the marketplace, the explanatory variables have little variability and they are usually highly collinear, which makes it difficult to obtain significant estimation results, usually, SP data is less costly to collect and less time consuming than gathering RP data. Clearly, a challenge of stated preference surveys is their hypothetical nature; consumers may provide unrealistic answers if there is no cost to overstating their willingness to pay. Also, if consumers are unfamiliar with the product (which could be the case for a new functional food attribute) their stated willingness to pay may be inaccurate, because the SP method asks respondents to state their preference values, but does not record an actual choice action as is the case with revealed preference studies. The fact that the SP method is based on what people say rather than what people do is the source of its greatest strengths and its greatest weaknesses. In empirical studies, researchers often use "cheap talk" to partially overcome the overstatement limitations. The "cheap talk" technique means that before the respondents complete the questionnaires, there are several statements to describe the product in detail by its attributes and encourage the respondents to imagine that they are in real consumption environments. When completing the survey questionnaires, some respondents may be inclined to provide responses that reflect broader notions of public welfare, rather than responding in a way that truly reflects their personal preferences, which might lead to biased responses. To some degree, using the "cheap talk" method encourages the 9 respondents to assume that they are in real shopping environments, and to provide responses that reflect their own behavioural choices. The stated preference method with discrete choice analysis is becoming increasingly popular among researchers. Larue et al Mixed Logit Model with heterogeneity of consumer preferences McFadden and Train According to that paper, the general random utility framework can be written as in formula (1). U ij is the consumer i's utility of product j. The variables (var1…varv) are the attributes of the product under study, with β's as parameters to be estimated and e ij as an unknown error term. U ij = ß 1 *var1 + ß 2 *var2 +...+ß v *varv + e ij (1) With some basic assumptions, the probability of the consumer i choosing different products j can be expressed in a familiar conditional logit form. However, unlike the traditional conditional logit models, which ignore these variations around the coefficients, the central idea of the mixed logit model is that consumers' choices are assumed to be heterogeneous through the specification distribution of the coefficients, which captures the consumer's heterogeneity preference Applying this approach to the case of functional food, the crucial part is to specify the determinant variables (different attributes related to functional food) in the utility function by which consumers make their consumption decisions. Potential variables include: (1) some health related variables, such as certain functions that consumers expect to get by choosing functional food; (2) some psychology variables, such as motivation and perception; (3) some demographic variables, and; (4) other factors, suc